System and method for applying ranking svm in query relaxation

Abstract

An enterprise-wide query relaxative support vector machine ranking algorithm approach provides enhanced functionality for query execution in a heterogeneous enterprise environment. Improved query results are obtained by adjusting ranking functions using machine learning methods to automatically train ranking functions. The improved query results are obtained using a list of document-query pairs that are modeled as a binary classification training problem, combination function which requires ranking and learning functions to be implemented representing document attributes and metadata utilizing query relaxation techniques and adjusted ranking functions. Machine learning methods implement user feedback to automatically train ranking functions.

Description

CROSS REFERENCE TO RELATED APPLICATIONS[0001]This application is related to U.S. patent application Ser. No. 10 / 434,845, filed Mar. 8, 2003, which is hereby incorporated herein by reference.COPYRIGHT NOTICE[0002]A portion of the disclosure of this patent document contains material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.BACKGROUND OF THE INVENTION[0003]The present invention relates generally to generating query results, as well as to information management systems that can be used with a heterogeneous enterprise environment, which can include structured data in a relational database as well as unstructured data stored in document images and document management applications. Embodiments also relate to applying ranking query results su...

AI Technical Summary

Benefits of technology

The present invention provides a system and method for improving search results for users, clients, or applications by using a support vector machine with query relaxant ranking. This approach involves obtaining ranked query item pairs, each pair corresponding to a query and a hit document, and calculating a relevancy score for each item pair based on a unique ranking function. The system can execute a series of query relaxation operations, allowing users to adjust the weighting functions and update the learned ranking function to improve the accuracy and usefulness of search results. The ranked query item pairs can be obtained from across a heterogeneous enterprise network, and the features can include an occurrence of a query term in metadata for a hit document, occurrence of a query term in a document body, and a document static rank. The system can be used in enterprise settings and can provide more accurate and useful search results for users.

Problems solved by technology

However, when an enterprise corpora consists of heterogeneous applications or where the same document has differing attributes emphasized in heterogeneous applications, keyword searches and drop down list searches of structured data will not meet the needs of a sophisticated enterprise search end user.

Large databases of documents, especially the World Wide Web, contain many “low quality” documents where the relevance to the desired search term is extremely low or non-existent.

As a result, searches typically return hundreds of irrelevant or unwanted documents that camouflage the few relevant documents that meet the personalized needs of an end user.

In other cases, however, these techniques are often not effective.

As the number of documents accessible via an enterprise intranet or the Internet grows, the number of documents that match a particular query becomes unmanageable.

Previous approaches for prioritizing searches have involved keyword priorities and pairs of keywords leading to some limited search results improvement.

As a result, a user may still be overwhelmed by an enormous number of documents returned by a search engine, unless the documents are ordered based on their relevance to the user's specific query and not merely limited to keywords or pairing of keywords.

Another problem is that differing deployments in a heterogeneous enterprise environment may want to emphasize different document attributes, creating a difficult task for a user attempting to return results from such a document.

However, in the context of differing attributes for the same document in a heterogeneous enterprise environment, such relevance ranking tools do not offer an end user the desired level of configurability and customization currently desired.

Ranking functions that rank documents according to their relevance to a given search query are known, and while useful in some settings, these functions do not allow a consistent user in a heterogeneous enterprise environment to personalize ranking results based on an end user set of preferences, either globally or for a single instance.

Images